In the Burn Injury Data described in Section 1.6.5 vital status at hospital discharge (DEATH) is the outcome variable. (a) Show that age (AGE) is not a confounder of the effect of inhalation injury...

In the Burn Injury Data described in Section 1.6.5 vital status at hospital discharge (DEATH) is the outcome variable.

(a) Show that age (AGE) is not a confounder of the effect of inhalation injury (INH_INJ) but is an effect modifier.

(b) Using the interactions model from part 5(a) and the four-step method prepare a table with estimates of the odds ratio and 95% confidence interval for inhalation injury for ages 20, 40, 60, and 80.

(c) Using the interaction model from part 5(a) prepare a graph of the estimate of the odds ratio for inhalation injury as a function of age.

(d) Add 95% confidence bands to the graph in part 5(c).

Settings where predicted probabilities are of interest tend be those where there is a reasonably wide range in the values. Conversely, if the range is too narrow graphs of fitted values tend to look like straight lines and thus are not much different, though on a different scale, than plots of fitted logits shown in Section 3.5 and add little to the analysis. Among the data sets described in Section 1.6 the Burn Study (Section 1.6.5) has the widest range of fitted values and we use it for the example in this section.